`structuredClone` parity questions

[syntacion]

While looking through Web Platform Tests's test suite for structured clones, I noticed a few more differences between msgpackr's implementation and the specs:

1. -0: -0 is separate from 0. We can easily add a check, but maybe we also don't want to unnecessarily bloat it to 5 bytes?
2. __proto__ keys: msgpackr silently replaces them with __proto_, presumably to avoid prototype pollution. I wonder if it might be better to use Object.defineProperty to set it instead? JSON.stringify and JSON.parse is able to handle these keys without issues, so I think msgpackr should try to do the same.
3. sparse array: Currently entirely unsupported by msgpackr.
4. sliced TypedArrays: If multiple TypedArrays reference the same buffer, the buffer is only used once. More strangely, structuredClone(new Uint8Array(1000000).subarray(0, 100)).buffer.byteLength === 1000000.

Do you think it makes sense to try to match structuredClone in each of these areas? Maybe put them behind an option?

[kriszyp]

1. Yeah, adding another byte doesn't worth the benefit.
2. So you would just use defineProperty for proto? That makes sense, I like that. And in https://github.com/kriszyp/msgpackr/blob/master/unpack.js#L521 we could reset the readObject.count so it never tries to create an object literatl.
3. Having an option for doing the (probably expensive) sparse array checking, makes sense.
4. This is a bug?